Inside two weeks of enterprise AI buying conversations. What buyers are actually asking for, and why the market hasn’t caught up.
“Integration is the biggest rock in everybody’s shoe.”
That was a CEO at a fintech infrastructure company on a discovery call last Thursday. He’s not the only one saying it. Across nine recent conversations with five enterprise AI buyers, the same themes keep showing up across every call. If you’re a prospective buyer talking to AI vendors right now, the gap between what most are pitching and what buyers are actually asking for is wider than it looks.
Here is what we’re hearing.
1. The legacy system is the focus.
Every buyer named their own legacy stack as the friction. Not their data quality, not their team’s AI literacy, not their budget. The stack. One described their core platform as “notoriously bad,” “outdated 90s-era UI, Windows-only,” with a “100-click site creation flow that’s embarrassing to train people on.”
These buyers have been at the integration problem for years. Rip-and-replace is off the table; their teams have already tried, failed, or been told no by procurement. What they want is a wrapper, a translator, an intelligence layer. Something that sits on top of what they already have and gets useful work out of it.
The vendors winning these conversations are not pitching transformation. They are pitching “we’ll ride alongside your legacy system.”
2. The schedule risk is on the buy-side, not the vendor’s.
In every active engagement we have right now, the timeline gate is heavily buyer-side. API keys waiting on procurement. Legal review pending on data ingestion. Sandboxes not yet stood up. Data residency decisions unresolved. One client has an internal standoff with their own platform vendor over upgrade costs that’s blocking everything downstream.
Your AI partner’s velocity does not matter if your access checklist isn’t done. The right delivery story is not “we ship fast.” It’s “we ship fast once your access is ready, and here is the checklist of what we need from you to make that happen in 30 days.”
If your vendor can’t hand you a pre-engagement access checklist on day one, that’s a signal.
3. The scoping conversation is where trust gets built. Or lost.
Pre-sales work is creeping. Buyers now expect vendors to invest in scoping the problem before the buyer invests budget in solving it. NDA first, technical deep dive with the co-founder, then a commercial conversation. That sequence has become more common.
It’s a fair ask. But the way a vendor handles it tells you almost everything about what working with them will be like. Vendors who give you a named, time-boxed scoping engagement with specific deliverables will likely run the production engagement the same way. Vendors who let “free assessment” stretch out for weeks without a meter on it are showing you how the contract will run.
4. Multi-party AI deals layer commercial overhead on top of product complexity.
A three-way deal we’re in right now: us, a strategic partner, and a hyperscaler, going through a federal procurement vehicle that adds a 7 to 8 percent markup. Initial pricing covers build cost only; margin gets deferred to subsequent phases. Pricing the work is one negotiation. The teaming agreement is another. The procurement vehicle is a third.
If you’re evaluating a multi-party AI engagement, the question to ask is the one every primary vendor is hoping you won’t: who actually owns the relationship 18 months in? Most multi-party deals don’t have a clean answer. That’s where they unwind.
What this means if you’re a buyer. The vendors you should be talking to lead with your legacy stack and your access bottlenecks, then their methodology. Ask them what their pre-engagement checklist looks like. Ask them what their scoping engagement produces. Ask them how they handle multi-party teaming when it exists. Most won’t have a clean answer to all four. The ones who do are the ones to keep talking to.
The pattern across these calls is not pessimistic. It’s clarifying. Buyers are getting better at knowing what they want. But the market hasn’t caught up. If you’re inside one of these conversations and the vendor in front of you is still pitching transformation, you have a signal.
If you’d like to talk through what your access checklist or scoping engagement should look like for your organization, get in touch.
Tom Burg leads B2B strategy at Turing Labs, working with marketing and revenue teams on the AI systems that turn disconnected signals into decisions. He writes about what it actually takes to make AI work inside real organizations: not in theory, but in production.